SuSE 11, mdadm and raid1

bbot · 08-27-2012, 12:23 PM

Hello.

Can I resync raid1 with rewrite 0x00 data that can't be read from dirty sectors at the end of the source disk?

Or I can do this by use dd only?

Raid1 was built with mdadm (mdadm build) and doesn't contain a superblock, so dd seems like a correct solution.

/dev/random · 08-30-2012, 02:40 PM

When you say dirty sectors, do you mean bad sectors?
If there are too many bad sectors on the disk not even dd can help you, sure you can tell it to copy the data over to another drive but any bad sectors it trys to read (conv=noerror,sync) will be replaced by zero's with enough damage done, there is no cheap way to reconstruct the data on that drive.

bbot · 09-05-2012, 03:11 AM

Yes, I mean bad sectors.

I did recovery with next commands.

Code:

#> mdadm -S /dev/md/raid1
#> sync && echo 3 > /proc/sys/vm/drop_caches

# Check write operations
#> iostat -x /dev/dm-* 1

First iteration ddrescue-1.16
#> ddrescue --force -n /dev/dm-1 /dev/dm-2 rescued.log
Second with no retry and smaller cluster-size
#> ddrescue --force -d -r0 --cluster-size 16 -A /dev/dm-1 /dev/dm-2 rescued.log
And last two tries to read bad sectors
#> ddrescue --force -d -r3 --retrim /dev/dm-1 /dev/dm-2 rescued.log
#> ddrescue --force -d -r10 --retrim /dev/dm-1 /dev/dm-2 rescued.log

#> sync && echo 3 > /proc/sys/vm/drop_caches

After that drives swapped and the raid array was rebuilded with no problems.